If stem cell transplant is not feasible, M.D. Anderson patients with relapsed AML typically receive investigational agents as first "salvage therapy," particularly if their initial remission has lasted less than one year. Although new agents are available, we do not know how a patient's AML will respond. In this program project, we will study therapies for which a "target" has been proposed, enabling us to examine relationships between clinical response and the status of the putative target (Specific Aim 1). Patients will be randomized, and we will assess their responses as being predictive of the target as well as study the use of biologic markers to assign patients to specific therapies. A confounding difficulty with Specific Aim 1 is the rarity of CR in relapsed AML. This rarity has motivated new definitions of response (e.g. CR p), which are less demanding than the criteria for CR. Preliminary data in untreated patients suggest that, while CR is essential for cure, patients who achieve CR p live longer than patients who achieve neither CR nor CR p after accounting for the time needed to achieve these responses. These data, and the low rate of CR p in the patients to be studied here, have prompted a plan to continue initial therapy in patients as long as they have neither clinical complications nor a rising peripheral blast count. This approach will allow us to compare the effect of various prospectively-defined responses on survival (Specific Aim 2). Two of the agents we will investigate (GX15-070MS, an inhibitor of the BH3 binding domain of Bcl-2 family members, and the kinase inhibitior BAY43-9006) will be given without chemotherapy; we will refer to these as "non-cytotoxic" therapies. The other two agents [the XIAP antisense oligonucleotide AEG35156 and AMD3100, an inhibitor of CXCR4 and thus of the blast cell-marrow stroma interaction), while themselves also non-cytotoxic, will be combined with idarubicin + ara-C (IA); these combinations will thus be called "cytotoxic." We will compare survival in: (a) patients given one of the two cytotoxic combinations only after an unsuccessful trial of one of the two non- cytotoxic agents, (b) patients in whom the alternate approachwas used and (c) patients given IA alone in previous studies. This approach will allow us to test the hypothesis that it appropriate to give some patients with relapsed AML non-cytotoxic therapy as initial treatment for relapse (Specific Aim 3). While such practice is increasingly widespread both in relapsed and untreated AML, its effect on survival remains unknown. Demonstration that, in at least one case, the obvious appeal of this strategy is not offset by a decrease in survival time would presumably be reassuring to patients.